Advances in High-Performance Computing and Electronics

The field of high-performance computing and electronics is rapidly evolving, with a focus on improving performance, reducing power consumption, and increasing scalability. Recent research has explored new architectures, such as 3D electronic-photonic heterogeneous interconnect platforms, which offer improved bandwidth density and energy efficiency. Additionally, advances in fabrication technology have enabled the development of nanomodular electronics, which promise to revolutionize the design and manufacturing of electronic circuits. Other areas of research include the optimization of phase-scheduling in digital circuits, the development of new memory technologies, and the improvement of collective communication algorithms for distributed machine learning applications. Noteworthy papers in this area include: NetCAS, which achieves up to 174% higher performance than traditional caching in remote storage environments. Algorithmic Tradeoff Exploration for Component Placement and Wire Routing in Nanomodular Electronics, which demonstrates significant optimization headroom in different dimensions, such as a $108 imes$ improvement in end-to-end manufacturing time. Short-circuiting Rings for Low-Latency AllReduce, which challenges the long-held assumption that the Ring algorithm is optimal only for large messages and presents a simple and fast heuristic for circuit-switching. Optimizing Phase-Scheduling with Throughput Trade-offs in AQFP Digital Circuits, which presents the first clock phase scheduling algorithm that combines phase-skipping and phase-alignment and achieves on average 6.8% area savings with a 2.62x increased throughput. Toward Co-adapting Machine Learning Job Shape and Cluster Topology, which demonstrates that both network contention and cluster utilization can be optimized simultaneously. 3D Electronic-Photonic Heterogenous Interconnect Platforms Enabling Energy-Efficient Scalable Architectures For Future HPC Systems, which proposes a 3D chiplet stacking electronic-photonic interconnect platform that offers a solution to the scaling issues of interconnect bandwidth and the memory wall problem. An Early Exploration of Deep-Learning-Driven Prefetching for Far Memory, which presents Memix, a far-memory system that embodies a deep-learning-system co-design for efficient and accurate prefetching. CMOS 2.0 - Redefining the Future of Scaling, which proposes to revisit the functional scaling paradigm by capitalizing on two recent developments in advanced chip manufacturing. On-Package Memory with Universal Chiplet Interconnect Express (UCIe), which proposes to enhance UCIe with memory semantics to deliver power-efficient bandwidth and cost-effective on-package memory solutions. Evaluating Rapid Makespan Predictions for Heterogeneous Systems with Programmable Logic, which aims to aid in the development of rapid makespan prediction algorithms by providing a highly flexible evaluation framework.

Sources

NetCAS: Dynamic Cache and Backend Device Management in Networked Environments

Algorithmic Tradeoff Exploration for Component Placement and Wire Routing in Nanomodular Electronics

Short-circuiting Rings for Low-Latency AllReduce

Optimizing Phase-Scheduling with Throughput Trade-offs in AQFP Digital Circuits

Toward Co-adapting Machine Learning Job Shape and Cluster Topology

3D Electronic-Photonic Heterogenous Interconnect Platforms Enabling Energy-Efficient Scalable Architectures For Future HPC Systems

An Early Exploration of Deep-Learning-Driven Prefetching for Far Memory

CMOS 2.0 - Redefining the Future of Scaling

On-Package Memory with Universal Chiplet Interconnect Express (UCIe): A Low Power, High Bandwidth, Low Latency and Low Cost Approach

Evaluating Rapid Makespan Predictions for Heterogeneous Systems with Programmable Logic

Built with on top of